GMDH神经网络在电主轴热位移建模中的应用  

Application of GMDH Neural Network in Thermal Displacement Modeling of Moterized Spindle

作　　者：万正海 李锻能[1] 潘岳健 WAN Zheng-hai;LI Duan-neng;PAN Yue-jian(School of Electro-Mechanical Engineering,Guangdong University of Technology,Guangzhou 510006,China)

机构地区：[1]广东工业大学机电工程学院

出　　处：《组合机床与自动化加工技术》2019年第6期9-11,16,共4页Modular Machine Tool & Automatic Manufacturing Technique

摘　　要：针对机床电主轴在高速运转时内部发热造成的热误差问题,对比BP、RBF神经网络方法,采用一种基于GMDH神经网络的电主轴热误差建模方法。以某型号高速数控机床电主轴为研究对象进行热误差实验,通过利用温度传感器和电涡流位移传感器分别采集主轴温度和轴向热位移数据,运用数据处理群集方法(GMDH)建立主轴轴向热误差预测模型。经过数据对比表明:该方法较传统的神经网络方法具有学习速度快、获得全局最优解、泛化性能好、拟合预测精度高等优点。Aiming at the thermal error caused by the internal heating of the machine tool spindle during high-speed operation,compared with the BP and RBF neural network methods.With the neural network based on GMDH motorized spindle thermal error modeling method,the thermal error experiment was carried out with a high-speed CNC machine tool spindle as a research object,using temperature sensor and eddy current displacement sensor collect spindle temperature and axial displacement data,use of data handling (GMDH) cluster method to establish the main shaft axial thermal error prediction model.Compared with the traditional neural network method,this method has the advantages of fast learning speed,optimal global solution, good generalization performance and high accuracy of fitting prediction.

关 键 词：数据处理群集方法(GMDH) 电主轴 热误差建模 轴向热位移 

分 类 号：TH161[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]